# LiteLLM — Self-hosted open-source gateway that puts models, agents and MCP tools behind one API. > Source: The Agents Index — https://theagentsindex.com/litellm (structured, researched, re-verified) LiteLLM is an open-source AI gateway and Python SDK for routing model, agent and MCP requests through an OpenAI-compatible interface. It is self-hosted, letting platform teams apply access controls, budgets and routing across their AI infrastructure. **Best for:** Platform teams operating production AI across models, agents and MCP tools | Fact | Value | | --- | --- | | Website | https://litellm.ai/ | | Category | LLM gateways & routers | | Open source | No | | Price | Free | | Price source | https://www.litellm.ai/pricing | | Price checked | 29 October 2026 | | MCP server | Yes | | API | Not stated | | Self-host | Yes | | Facts | 21 facts from 12 sources | | Facts checked | 29 October 2026 | Screenshot of LiteLLM: https://theagentsindex.com/img/1920/screenshots/litellm.webp, captured 31 October 2026 **What we checked** - Pages last read: 29 October 2026 - Facts checked: 21 facts (From 12 sources) - AI answers: Named in 0 of 151 (Our October check) ## What we checked on LiteLLM ### How much it does unattended **Agent permissions** · Yes — The gateway controls MCP tool access by key and team, so an agent receives only permitted tools. > You control access to each tool by key and by team. An agent gets only the tools that you permit. > — https://www.litellm.ai/about · MCP gateway section, checked 29 October 2026 ### Which models it runs on **Open models** · Yes — LiteLLM supports locally run and self-hosted model providers including Ollama, vLLM, LM Studio, and Llamafile. > ## Self-Hosted & Local - Ollama - vLLM - LM Studio - Llamafile > — https://docs.litellm.ai/docs/providers · Self-Hosted & Local provider list, checked 29 October 2026 **Model choice** · Yes — Buyers select a provider-prefixed model name, while the same request format works from both the Python SDK and gateway. > Set the model as `/`, for example `anthropic/claude-sonnet-5`, and the same request works from the Python SDK and the Gateway. > — https://docs.litellm.ai/docs/providers · AI Providers introduction, checked 29 October 2026 ### Whose machine it runs on **Self-hosted** · Yes — The free gateway is intended for self-hosting, and the vendor also documents air-gapped deployment for Enterprise customers. > The open-source gateway is free to self-host, forever. > — https://www.litellm.ai/pricing · Pricing hero, checked 29 October 2026 **Open-source runtime** · Yes — A buyer can run the open-source gateway indefinitely on its own infrastructure without needing a LiteLLM account or license fee. > Yes. Self-host LiteLLM OSS with no license fee, forever. > — https://www.litellm.ai/pricing · Pricing FAQ, Is the open-source version really free for production?, checked 29 October 2026 **Runs in your cloud** · Yes — The gateway supports deployment in a buyer's AWS, Google Cloud, Azure, or Kubernetes environment using Helm and Terraform paths. > Production deployment guide for LiteLLM on AWS, Google Cloud, Azure, or any Kubernetes cluster. > — https://docs.litellm.ai/docs/proxy/deploy · Page introduction, checked 29 October 2026 ### Building with it **SDK languages** · Yes — Python is LiteLLM's native SDK language; non-Python applications access the gateway through OpenAI-compatible client interfaces. > Python SDK: `uv add litellm` > — https://docs.litellm.ai/docs/ · Getting Started, Quick Start: start the LiteLLM Gateway, checked 29 October 2026 **Tracing and evals** · Yes — Each request gets a unique call identifier and can send input, output, and exception information to supported logging systems. > LiteLLM generates a unique `call_id` for each request. This `call_id` can be used to track the request across the system. > — https://docs.litellm.ai/docs/proxy/logging · Getting the LiteLLM Call ID, checked 29 October 2026 ### What it costs to run **How it meters** · Partly — The OSS gateway has no license fee, while Enterprise is an annual quote based on request capacity, deployment architecture, and support needs. > Enterprise pricing is sized to your annual gateway request capacity, deployment architecture, and support needs — never per token. > — https://www.litellm.ai/pricing · Enterprise pricing section, checked 29 October 2026 **Cost of one job** · Partly — LiteLLM adds no gateway fee to model spend, but it does not publish a per-job price because underlying model providers bill separately. > LiteLLM adds no fee to your model cost. > — https://www.litellm.ai/about · What makes LiteLLM different, Open source, with no fee on tokens, checked 29 October 2026 **Free tier** · Yes — The open-source gateway is a standing free tier, self-hosted with no license fee, covering virtual keys, budgets, spend tracking, fallbacks and Prometheus metrics; a separate 30-day Enterprise trial key exists on top. > Is the open-source version really free for production? Yes. Self-host LiteLLM OSS with no license fee, forever. > — https://www.litellm.ai/pricing · Pricing FAQ, question 01, checked 28 October 2026 **MCP server** · Yes — The gateway acts as an MCP gateway: it fronts MCP servers over streamable HTTP, SSE and stdio, lists and calls their tools, and scopes access by key, team or organization. stdio needs a flag set at startup. > LiteLLM Proxy provides an MCP Gateway that allows you to use a fixed endpoint for all MCP tools and control MCP access by Key, Team. > — https://docs.litellm.ai/docs/mcp · MCP Overview, opening paragraph and Overview table, checked 28 October 2026 **Bring your own key** · Yes — Deployments retain provider keys in their own infrastructure instead of sharing a vendor-managed inference credential. > Nobody shares provider keys. > — https://www.litellm.ai/about · AI Gateway section, checked 29 October 2026 ### Buying it for a team **A company can buy it** · Yes — Organizations can obtain an Enterprise license directly or through cloud marketplaces and authorized resellers for production deployments. > Direct, or through AWS Marketplace and authorized resellers if that fits your procurement. > — https://www.litellm.ai/pricing · Pricing FAQ, How do we buy?, checked 29 October 2026 **Seat model** · Partly — The vendor publishes no per-seat price; it offers SSO free for up to five users and requires Enterprise licensing beyond that threshold. > SSO is free for up to 5 users; beyond that, an Enterprise license is required. > — https://docs.litellm.ai/docs/enterprise · Who is Enterprise for?, checked 29 October 2026 **Pooled budget** · Yes — Team virtual keys can draw from a shared team budget, while optional member budgets set individual limits inside that pool. > Team budgets: Add team_id to virtual keys to draw on a team's shared budget > — https://docs.litellm.ai/docs/proxy/users · Budget Setup Options, checked 29 October 2026 **Admin controls** · Yes — Administrators can assign scoped organization and team roles, then constrain budgets, rates, models, and member permissions. > Team admins manage keys, models, and budgets per project, tag, and model without the master key > — https://docs.litellm.ai/docs/enterprise · What changes for your organization, Teams run their own keys and budgets, checked 29 October 2026 **Audit log** · Partly — Enterprise records creation, updates, deletion, and regeneration of keys, teams, users, and models, rather than exposing this as an OSS entitlement. > LiteLLM Enterprise feature. Requires an Enterprise license (`LITELLM_LICENSE`). > — https://docs.litellm.ai/docs/proxy/audit_logs · Audit Logs introduction, checked 29 October 2026 **Single sign-on** · Partly — The Admin UI supports SSO for up to five users without charge, while larger deployments require an Enterprise license. > From v1.76.0, SSO is free for up to 5 users. Beyond that, an enterprise license is required. > — https://docs.litellm.ai/docs/proxy/admin_ui_sso · SSO for Admin UI introduction, checked 29 October 2026 ### What happens to your code **Data residency** · Yes — Self-hosted deployments keep processed data in the buyer's infrastructure, so the chosen deployment location determines residency. > Any data generated or processed remains entirely within your own infrastructure. > — https://docs.litellm.ai/docs/data_security · Collection of Personal Data, For Self-hosted LiteLLM Users, checked 29 October 2026 **Certifications** · Yes — LiteLLM states that it is SOC 2 Type II audited, with the current report available through its Trust Center. > SOC 2 Type II. Our current report is available through the LiteLLM Trust Center. > — https://docs.litellm.ai/docs/data_security · Legal/Compliance FAQs, Vendor Information, checked 29 October 2026 ## LiteLLM price and terms **Terms** - Free plan: Yes — Open-source, self-hosted gateway — free forever with no license fee. Underlying model-provider usage is separate. **What needs a higher plan** - Enterprise: Audit log ## For AI agents - MCP server: Yes Pricing page as we captured it: https://theagentsindex.com/img/1920/screenshots/litellm/pricing.webp, captured 31 October 2026 Read on https://www.litellm.ai/pricing, 29 October 2026. ## LiteLLM alternatives - [OpenRouter](https://theagentsindex.com/openrouter) · Free plan: One OpenAI-compatible API that routes across hundreds of models and providers. - [Portkey](https://theagentsindex.com/portkey) · Free plan, then $49 a month: An open-source AI gateway that unifies model access with governance and LLM observability. All 2 publish a price. [All 2 alternatives](https://theagentsindex.com/litellm/alternatives/) ## Questions about LiteLLM ### How much does LiteLLM cost? LiteLLM is free to use. From litellm.ai/pricing, checked 29 October 2026. ### Does LiteLLM have an API or an MCP server? It has an MCP server. ### Is LiteLLM open source, and can you host it yourself? You can host it yourself. ### What have we checked on LiteLLM? 21 facts, under these headings: How much it does unattended, which models it runs on, whose machine it runs on, building with it, what it costs to run, buying it for a team, what happens to your code. ### How do we check LiteLLM? We hold 21 facts on LiteLLM, quoted from 12 pages. Facts and price last checked 29 October 2026. How we check every listing: https://theagentsindex.com/methodology/ ## Sources and changes Last changed 31 October 2026 **Sources** (12) - https://www.litellm.ai/pricing, read 29 October 2026 - https://www.litellm.ai/about, read 29 October 2026 - https://docs.litellm.ai/docs/data_security, read 29 October 2026 - https://docs.litellm.ai/docs/enterprise, read 29 October 2026 - https://docs.litellm.ai/docs/providers, read 29 October 2026 - https://docs.litellm.ai/docs/, read 29 October 2026 - https://docs.litellm.ai/docs/mcp, read 28 October 2026 - https://docs.litellm.ai/docs/proxy/admin_ui_sso, read 29 October 2026 - https://docs.litellm.ai/docs/proxy/audit_logs, read 29 October 2026 - https://docs.litellm.ai/docs/proxy/deploy, read 29 October 2026 - https://docs.litellm.ai/docs/proxy/logging, read 29 October 2026 - https://docs.litellm.ai/docs/proxy/users, read 29 October 2026 ## Is this yours? AI assistants like ChatGPT and Claude opened The Agents Index pages 2,746 times in September. Claim LiteLLM: https://theagentsindex.com/claim/litellm Report an error if a fact is wrong: hello@theagentsindex.com, subject "Report an error: LiteLLM (litellm)".